Kubernetes YAML Manifests
Introduction
So far you've used imperative commands (kubectl create, kubectl expose). For real work,
you write YAML manifests that describe your desired state. This is the declarative
approach that makes Kubernetes powerful and GitOps possible.
Why This Matters
Imperative commands are great for learning and debugging, but terrible for production:
- How do you recreate what you did?
- How does your teammate know what's deployed?
- How do you review changes before applying them?
- How do you roll back?
YAML manifests solve all of this. Your Kubernetes configuration becomes code that's
versioned in Git, reviewed in PRs, and applied through CI/CD.
In MLOps, this is how you'll define:
- Model serving deployments
- Training job specifications
- Monitoring and observability stack
- The entire ML platform infrastructure